The Agent Skills Directory

[METADATA_POISONING]: The SKILL.json file contains hardcoded absolute file paths (e.g., /home/toowired/.claude-memories/index.json) that appear to be specific to the author's local development environment. This contradicts the generic, platform-agnostic path instructions (using ~ and %USERPROFILE%) provided in the SKILL.md file, making the metadata deceptive or misleading for general use.
[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external, potentially untrusted sources and interpolates it into agent outputs without sanitization.
Ingestion points: The skill reads from ./.context/current.json, ./.context/session.json, ./DECISIONS.md, ./README.md, and ~/.claude-memories/index.json (as specified in SKILL.md).
Boundary markers: There are no instructions or delimiters defined to prevent the agent from following malicious commands embedded within these files.
Capability inventory: The skill has the ability to execute shell commands (e.g., git), write session state files, and generate HTML artifacts.
Sanitization: The skill lacks any mechanism to escape, filter, or validate the content of memories or project files before they are included in the session summary or the visual HTML dashboard.

session-launcher